DEgenes Hunter - Differential expression analysis report
Details of the input data
First group of samples (to be referred to as control in the rest of the report)
| Sample Names: |
|---|
| CTL_1 |
| CTL_2 |
| CTL_3 |
| CTL_4 |
Second group of samples (to be referred to as treatment in the rest of the report)
| Sample Names: |
|---|
| epm2a_1 |
| epm2a_2 |
| epm2a_3 |
| epm2b_1 |
| epm2b_2 |
| epm2b_3 |
| epm2b_4 |
Note: A positive log fold change shows higher expression in the treatment group; a negative log fold change represents higher expression in the control group.
Data quality control (QC)
Correlation between samples:
Here we show scatterplots comparing expression levels for all genes between the different samples, for i) all controls, ii) all treatment samples and iii) for all samples together.
These plots will only be produced when the total number of samples to compare within a group is less than or equal to 10.
Correlation between control samples:
Replicates within the same group tend to have Pearson correlation coefficients >= 0.96. Lower values may indicate problems with the samples. 
Correlation between treatment samples:
Replicates within the same group tend to have Pearson correlation coefficients >= 0.96. Lower values may indicate problems with the samples. 
Heatmap and clustering showing correlation between replicates
BROWN: higher correlation; YELLOW: lower
Principal Component Analysis
This is a PCA plot of the count values following rlog normalization from the DESeq2 package:
The samples are shown in the 2D plane and distributed by their first two principal components. This type of plot is useful for visualizing the overall effect of experimental covariates and batch effects. It is also useful for identifying outlier samples. Control and treatment samples respectively may cluster together.
Visualizing normalization results
These boxplots show the distributions of count data before and after normalization (shown for normalization method DESeq2):
Before normalization:
After normalization:
Samples differences by all counts normalized:
All counts were normalizated by DESeq2 algorithm. This count were scaled by log10 and plotted in a heatmap.
DEgenes Hunter results
Gene classification by DEgenes Hunter
DEgenes Hunter uses multiple DE detection packages to analyse all genes in the input count table and labels them accordingly:
- Filtered out: Genes discarded during the filtering process as showing no or very low expression.
- Prevalent DEG: Genes considered as differentially expressed (DE) by at least 3 packages, as specified by the
minpack_commonargument. - Possible DEG: Genes considered DE by at least one of the DE detection packages.
- Not DEG: Genes not considered DE in any package.
This barplot shows the total number of genes passing each stage of analysis - from the total number of genes in the input table of counts, to the genes surviving the expression filter, to the genes detected as DE by one package, to the genes detected by at least 3 packages.
Package DEG detection stats
This is the Venn Diagram of all possible DE genes (DEGs) according to at least on of the DE detection packages employed:
FDR gene-wise benchmarking
Benchmark of false positive calling (Image extracted from {“padj_prevalent_DEGs.pdf”} file):
Boxplot of FDR values among all genes with an FDR <= 0.05 in at least one DE detection package
The complete results of the DEgenes Hunter differential expression analysis can be found in the “hunter_results_table.txt” file in the Common_results folder
DE detection package specific results
Various plots specific to each package are shown below:
DESeq2 normalization effects:
This plot compares the effective library size with raw library size
## `geom_smooth()` using formula 'y ~ x'
The effective library size is the factor used by DESeq2 normalizatioin algorithm for eahc sample. The effective library size must be dependent of raw library size.
DESeq2 MA plot:
This is the MA plot from DESeq2 package:
In DESeq2, the MA-plot (log ratio versus abundance) shows the log2 fold changes are attributable to a given variable over the mean of normalized counts. Points will be colored red if the adjusted Pvalue is less than 0.1. Points which fall out of the window are plotted as open triangles pointing either up or down.
A table containing the DESeq2 DEGs is provided: in Results_DESeq2/DEgenes_DESEq2.txt
A table containing the DESeq2 normalized counts is provided in Results_DESeq2/Normalized_counts_DESEq2.txt
Differences between samples by PREVALENT DEGs normalized counts:
Counts of prevalent DEGs were normalizated by DESeq2 algorithm. This count were scaled by log10 and plotted in a heatmap.
edgeR MA plot
This is the MA plot from edgeR package:
Differential gene expression data can be visualized as MA-plots (log ratio versus abundance) where each dot represents a gene. The differentially expressed genes are colored red and the non-differentially expressed ones are colored black.
A table containing the edgeR DEGs is provided in Results_edgeR/DEgenes_edgeR.txt
A table containing the edgeR normalized counts is provided in Results_edgeR/Normalized_counts_edgeR.txt
limma Volcano plot
Volcano plot of log2-fold change versus -log10 of adjusted p-values for all genes according to the analysis with limma:
A table containing the limma DEGs is provided in Results_limma/DEgenes_limma.txtA table containing the limma normalized counts is provided in Results_limma/Normalized_counts_limma.txt
WGCNA Results
WGCNA was run to look for modules (clusters) of coexpressed genes. These modules were then compared with the sample factors to look for correlation. If no sample factors were specified, this comparison was performed with treatment/control labels.
The following graphic shows the power value chosen for building clusters. The power is chosen by looking at the characteristics of the network produced.
In total there were 290 clusters. The following plot shows the number of genes per cluster:
The following plots show the correlation between the different modules and specified factors. This is done using eigengenes, which can be broadly thought of as the average expression pattern for the genes in a given cluster. MEn refers to the eigengene for cluster n.
Cluster assignment vs lower module membership (MM)
This plot shows, for each gene, the cluster ID ascertained by WGCNA, vs. the cluster whose eigen gene has the highest correlation (module membership/MM).
Cluster vs. factors correlation
This plot shows the correlation between clusters (eigen genes) and factors directly.